Explore every episode of the podcast From Models to Medicine

For this episode, we sit down with Kristeen Pareja-Navarro, co-CEO of Engram Bio, whose path from a farming family in the Philippines to peptide engineering for Alzheimer's disease at the Buck Institute is one of the more remarkable stories we've heard on the show. Kristeen walks us through her work trimming the KIBRA protein into a small functional peptide that rescues memory impairment in mice, and how tools like AlphaFold 3 and ESMFold have compressed years of structural biology work into days.

But she's equally candid about where AI falls short. Intrinsically disordered proteins still break the models. AI summaries miss the nuances buried in figures and methods. And no amount of compute can answer the question the field is still wrestling with: how does memory persist when every cell in your body eventually turns over. A grounded, honest conversation about what AI is genuinely changing in early-stage drug development and what it isn't.

Show Notes:

• https://www.biorxiv.org/content/10.64898/2026.06.03.729735v1.full.pdf

• https://www.nature.com/articles/s41586-025-09429-6

We sit down with Shawn Manchester, CEO of TripleBar, whose path from yeast DNA repair to metabolic engineering at MIT to biomanufacturing AI gave him a front-row seat to one of biotech's most underappreciated problems.

Everyone funds discovery. Nobody funds manufacturing. And yet a molecule that can't be made reliably and cheaply at scale never reaches a patient. Shawn makes the case that biomanufacturing is where the real bottleneck lives, and why it's finally getting the AI attention it deserves. It's a refreshingly grounded conversation about what biological AI actually requires to work.

In this episode of From Models to Medicine, we sit down with Angie Schwab, CEO and founder of Trialynx, who has spent 20 years watching clinical trials run over budget, over timeline, and over complexity for one reason: nobody got rid of the copy-paste. Angie makes a sharp distinction that most people in the AI conversation are blurring. That automation and AI are not the same thing, and deploying generative AI on top of unstructured workflows without fixing the foundation first is how you get polished-looking documents that still require someone to fact-check every line.

We get into how Trialynx compresses months of protocol documentation into 30 days, why a single source of truth changes everything downstream, and where Angie draws a hard line on what AI should never touch. The episode closes on a genuinely provocative idea: AI pseudo-clinical trials that run molecules through synthetic models of human biology before a single patient is enrolled.

This week, we sit down with Jean-Baptiste Morlot of Deep Life, who makes the case that the drug discovery field has been asking the wrong thing of its AI models. Correlation-based machine learning can find patterns, but it can't tell you why a cell is sick or what it would take to make it healthy again. Jean-Baptiste walks us through how single-cell sequencing unlocked the data volume needed to train something far more ambitious: a causal world model of cellular biology that learns the underlying rules of disease rather than memorizing them.

We get into how Deep Life's TwinCell platform works, why most virtual cell benchmarks are quietly inflated by popularity bias, and what it means to build an interpretable AI that a biologist can trust.

This week, we sit down with Katie Lennon, CSO of Exthymic, who makes a case most bench scientists aren't ready to hear: biology isn't just a discovery discipline anymore. It's engineering. Katie walks us through what it actually looks like to build biology, hardware, and software concurrently, why that requires a shared language across disciplines, and what gets lost when teams hand off instead of collaborate.

We also get into the unglamorous side of AI in the lab. More data documentation, peer-checking metadata before an experiment is marked complete, and knowing exactly where LLMs quietly get things wrong. Katie also shares a surprisingly effective strategy for getting scientists over their fear of AI tools.

Companies mentioned in the show:

https://exthymic.com/
https://oni.bio/
https://www.potato.ai/
https://ollie.ai/

-------------------

Sign up for KAMI Think Tank's Maven course here! Classes start in less than 10 days!!

In this episode of From Models to Medicine, we sit down with Mark Amouzgar, co-founder and CEO of March Health. Mark opens with a striking stat: general-purpose AI models are only about 60% accurate in women's health. Sounding confident isn't the same as being clinically safe, and in this episode he breaks down exactly why generic AI was never built for this.

We get into how March Health built a gender-aware digital twin, what it takes to wrap a model in real clinical logic, and why the average endometriosis patient sees eight different doctors before getting an accurate diagnosis. It's a candid conversation about a problem that doesn't get nearly enough attention.

-------------------------

Sign up for KAMI Think Tank's Maven course here!

In this episode of From Models to Medicine, we sit down with Lavanya Anandan, Head of External Innovation, West Coast at CSL, whose path from bench scientist to Merck Group platform builder to UCSF Innovation Venture startup advisor gives her a rare view of where AI in drug discovery is delivering versus where it's mostly promise. She makes a sharp case that the companies winning in AI-bio aren't necessarily the ones with the best models, and the reason why might surprise you.

We get into what big pharma wants from an AI partner, why women's health data scarcity is quietly producing some of the most rigorous AI in biology, and what an unexpected finding in mRNA research tells us about where this whole field might be heading.

Articles mentioned in show: https://endpoints.news/can-ai-do-scientific-research-billions-chase-hopes-of-superintelligence/

Maven course: https://maven.com/kamithinktank/ai-literacy-for-life-sciences-healthcare

We chat with Karl Moritz, CEO of Reliant AI, now part of Cohere, whose career took him from NLP research to McKinsey to building AI infrastructure for drug development. Karl reminds us that clinical trial success rates have been stuck years, and the golden age of AI-powered target discovery isn't fixing it.

Karl lays out a sharp framework; own the science, rent the AI plumbing, and explains why building your own data ingestion pipeline is as misguided as writing your own operating system. He's also candid about what general-purpose LLMs are actually good for, where purpose-built enterprise tools earn their keep, and why the audit trail question will separate serious AI vendors from everyone else. If you're a pharma or biotech leader trying to make sense of the build vs. buy debate right now, this one is worth your time.

For this episode, we sit down with Banishree Saha, Global Biomarkers Lead at Takeda, who has worked in fields such as academic liver biology to translational biomarker work, over the course of her 20-year career. She walks us through what biomarkers really are, how flow cytometry went from single-color panels to capturing 40 proteins simultaneously, and why that explosion in data has made systems thinking the only way forward.

Banishree talks about AI's strengths and weaknesses, and makes the case that the scientists most at risk from AI aren't the ones who know too much biology. But are the ones who don't know enough to catch when the model is wrong.

In this episode, we sit down with Chris Nelson, an mRNA scientist at nCode Bio whose career has taken him from building blood transfusion matching tests now used by the NHS, to cancer genomics at Personalis, to the frontier of mRNA sequence design. Chris breaks down why the untranslated regions of mRNA might actually change how we dose therapeutics, reduce side effects, and engineer cells inside the body entirely.

Then we talk about why AI has conquered protein folding but keeps hitting a wall with mRNA. Chris walks us through the data problem, the cost problem, and the modeling choices nCode Bio is making that go against the grain of what most of the field is building. He's also candid about where the big model providers are letting biology down and which scrappy research groups he thinks are quietly building the tools that will actually matter.

In this episode of From Models to Medicine, we sit down with Sarah Daniels, an environmental health researcher, who makes the argument that the chemicals making us sick might also be our best clues for building better drugs. She traces that idea from a 1976 discovery about dioxins all the way to an FDA-approved psoriasis treatment in 2023 — and asks what AI could accelerate if the field finally got its data act together.

Sarah brings up where AI in environmental health is genuinely stuck; siloed industry data, models built only for acute toxicity, and a chronic low-dose problem that no one has cracked. She also shares how she used APIs to build her own real-time funding tracker for microplastics research in under a month, and why the gap between what AI promises and what environmental health can actually feed it is one of the most under appreciated bottlenecks in the field right now.

Links mentioned in this episode:

1. Historic ARPA-H program funding announcement on Microplastics: https://arpa-h.gov/explore-funding/programs/stomp
2. Sarah's two cents on the ARPA-H Microplastics program:  https://engineeredresilience.substack.com/p/a-microplastics-prelude-for-engineered
3. Aggregated trends in microplastics from the non-peer-reviewed literature (Trendspotter App):https://www.engineeredresilience.org/microplastics
4. Stay up to date on latest releases from Sarah's work on Engineered Resilience by signing up for substack: https://engineeredresilience.substack.com/
5. Report on unexpected health effects observed after LA fires showing how questions remain post-exposure to pollutants:https://www.latimes.com/environment/story/2025-12-17/la-fires-heart-attacks-strange-blood-test-results-spiked
6. Can a Biologist Fix a Radio? - great piece on the necessity to understand biology at the systems-level: https://www.cell.com/cancer-cell/fulltext/S1535-6108(02)00133-2

Stan Jastrzębski is the co-founder of molecule.one and a deep learning researcher who made a deliberate pivot into synthetic chemistry. In this episode, he explains why chemistry is the real bottleneck in drug discovery today. Unlike biology, which has the Protein Data Bank and tools like AlphaFold, chemistry lacks the massive, balanced datasets AI needs to work. Scientific literature makes it worse, publishing wins and burying failures, which starves models of exactly the negative data they learn from most.

Stan also talks about "vibe binding," the industry's growing tendency to over-rely on biological binding models that only work on well-trodden targets and quietly kill scientific creativity in the process. He closes with a sharp take on where LLMs actually hit their ceiling, and why he thinks scientific discovery is not just a use case for AI but its ultimate test.

Anna Benefiel has spent over 20 years in healthcare innovation, and she's tired of watching life sciences sit on the sidelines of quantum computing. In this episode, she makes a direct case for why pharma needs to stop treating quantum as a future problem and start treating it as a present opportunity. We get into what quantum actually does that classical computing cannot, why probabilistic nuance matters for complex biological problems, and where the real near-term value lives: clinical trial optimization, supply chain, and cell culture yields.

We also dig into what happens when you stack quantum on top of AI. Anna points to models already using standard histopathology slides to generate spatial proteomic maps for lung cancer, and argues quantum is what takes discoveries like that to the next order of magnitude.

*The views represented on this podcast are not a reflection of Strangeworks.

In this episode of From Models to Medicine, we speak with Vid Stojevic, the co-founder and CEO of Kuano, a Cambridge-based company using quantum algorithms and AI to tackle the drug discovery problems that traditional computational chemistry keeps failing. In this episode, he explains why he deliberately ignored the broad platform play and went narrow instead, targeting the specific early-stage problems where getting the physics right changes everything.

We get into what a "quantum lens" actually means in practice, why transition states are a better design target than natural substrates, and how Kuano is succeeding on targets that pharma had written off as undruggable. Vid makes a sharp case for how generating synthetic quantum data turns a low-data drug discovery problem into something AI can actually work with. He closes with honest advice on when quantum simulation is the right tool and when it simply isn't.

Jenny Yang is the co-founder and CEO of Outpost Bio, where her team is working to make human microbiology computable. In this episode, she breaks down why bias in ML models is so easy to miss. High overall accuracy can hide terrible performance on specific subgroups, and in healthcare, that gap has consequences. She traces the problem upstream, from skewed training datasets to the way clinical definitions themselves carry historical bias, and explains the real trade-offs involved in trying to correct for it.

We also get into what makes the microbiome such a hard problem is that our microbiomes can differ by up to 90% from person to person. Jenny walks us through how Outpost Bio's "lab in the loop" model tightly integrates wet lab experiments with AI to generate better, less biased data from the ground up, and why rigorous external validation is the thing she'd tell every biotech founder to prioritize before anything else.

Bogdan Knezevic is the CEO and co-founder of Kaleidoscope Bio, and he's seen enough failed AI implementations to know where they almost always break down. In this episode, he walks us through what minimum viable data standardization actually looks like in practice, why consistent naming conventions and structured data entry matter more than people want to admit, and what every biotech CEO should ask their team before writing another AI budget line.

We also get into a guardrails conversation and Bogdan is direct about what happens when autonomous agents operate without proper permissions. He closes with a sharp framework for deciding what to build in-house versus hand off, with some great resource hand-offs.

• Data vs keys
• AI needs context
• Iterating your way to success via better data management
• The data maturity ladder
• Before AI there was good data

In this episode of From Models to Medicine, we sit down with Minna Schmidt, a postdoctoral researcher at the Buck Institute for Research on Aging. Minna walks us through Braak's hypothesis and the emerging "brain-first vs. body-first" framing of the disease, explaining how symptoms can appear up to 30 years before a clinical diagnosis is ever made.

We also get into the data side of the work. Minna uses a dataset with over 54,000 participants and talks honestly about what AI actually does and doesn't unlock when you're staring down that volume of microbiome and clinical data. She uses LLMs to organize her thinking, speed up literature reviews, and learn basic programming, while being clear-eyed about where the field's biggest bottleneck actually is: not the tools, but the data itself.

----------------------------------------------------------------------

This episode was sponsored by ⁠CleanSpace⁠.

⁠CleanSpace⁠ designs, manufactures, and installs advanced controlled environments—delivering complex projects months faster with guaranteed costs and uncompromising performance.

Please contact Chelsea for more information or with any questions at CLauridsen@CleanSpaceus.com.

In this episode of From Models to Medicine, we sit down with Sal Tejani, Associate Director for Field Medical Affairs at Regeneron*, who started his career catching dangerous prescription errors at CVS and never lost the instinct for finding the lever that actually moves things. Today that instinct is pointed squarely at AI; how to use it, when to trust it, and when it will absolutely get you into trouble.

Sal gives us an honest, practitioner-level view of what AI looks like inside a major pharma company: the tools that are actually useful, the guardrails that are non-negotiable, and the human judgment that no model has figured out how to replace yet. Plus, he closes with a personal story that reframes the whole conversation about why any of this actually matters.

This episode was sponsored by CleanSpace.

CleanSpace designs, manufactures, and installs advanced controlled environments—delivering complex projects months faster with guaranteed costs and uncompromising performance.

Please contact Chelsea for more information or with any questions at CLauridsen@CleanSpaceus.com.

*Thoughts brought up on this podcast do not represent the views of Regeneron.

Rachel Jacobson has spent her career moving between some of the most demanding corners of life sciences before founding Powerhouse Biology. In this episode, she traces that journey and explains why, after all of it, she keeps coming back to mitochondria. We get into what it actually takes to bridge biology and machine learning inside a lab culture, why asking "stupid questions" across disciplines is a feature and not a bug, and what she had to unlearn from traditional drug development to work effectively alongside ML engineers.

We also dig into data design. Rachel makes a sharp case that data passing standard biological QC is not the same as data that's ready for a machine learning model. Uneven plate layouts, cell debris, different scientists handling samples can all create batch effects that quietly break your model before it ever sees a hypothesis worth testing. She connects all of this to a bigger argument about why human biological variability needs to be built into preclinical pipelines from the start, and why ML might finally give scientists the tools to do that seriously.

In this episode, we sit down with Dr. Freddy Nguyen, CEO and co-founder of Nine Diagnostics, whose background spans medicine, pathology, optics, and nanotechnology. Freddy shares how Nine Diagnostics is building a multiomics platform that helps cancer patients find out within days whether their treatment is actually working.

We dig into why AI's real power in medicine lies in its ability to connect siloed data across molecular readouts, imaging, and clinical context, and why treating patients as more than just their diagnosis is the only way to build tools that actually hold up in the real world. We also get into the harder conversation: where AI in clinical workflows breaks down. It's a candid, technically grounded conversation about what equitable AI in medicine actually requires.

In this episode of From Models to Medicine, we sit down with Elisa Martin Perez, a postdoc at University of California, Berkeley, to talk about how non-coders are starting to use AI in their day-to-day work. From learning R through conversation to making sense of massive CRISPR screens, Elisa shares how AI is becoming a practical tool for navigating data, checking experimental design, and cutting down on the kinds of manual tasks that quietly consume hours in the lab.

We also get into the hesitation many scientists feel around adopting AI, where the technology actually helps (and where it doesn’t), and why it still falls short of running experiments end-to-end. Along the way, we touch on lab logistics, data overload, and what it means to use AI as a thinking partner rather than a replacement.

What does it actually take to make AI work inside a pharmaceutical company and why do so many efforts stall after the model is built? Pranay Mohanty, from J&J Innovative Medicine, joins us to talk about how building the model is only one part of the work and what can happen when you try to apply that model to messy, real-world data.

We dig into how teams are starting to use digital twins to simulate patients and rethink trial design, what it looks like to work alongside regulators like the U.S. Food and Drug Administration, and why keeping a human in the loop isn’t optional. Along the way, we share how AI is actually shaping portfolio decisions today and where the limits still are.

In this episode of From Models to Medicine, we sit down with Ashley Zehnder, the founder of Fauna Bio to explore one of the most creative bets in drug discovery: using extreme animal biology to unlock human therapeutics.

We dig into the graph neural networks powering Fauna Bio's drug discovery platform, the unglamorous data curation work that makes it all function, and why human-in-the-loop isn't a limitation but a design principle. Plus: what hibernation research could mean for long-duration human space travel.

What does emergency medicine have to do with large-scale knowledge systems? Jason Grafft joins us to talk about his non-linear path from EMS and simulation education to knowledge engineering and why the structured, fact-bound thinking that medicine demands is one of the most undervalued assets in the AI era. We dig into how he uses graph-based models to bridge clinical experts and technical systems, where LLMs are actually helping in entity resolution, and the practical advice he gives every clinician trying to break into tech.

In this episode of From Models to Medicine, we sit down with Rachel Thomas to discuss the growing gap between AI hype and scientific reality in the life sciences.

The conversation highlights the hidden costs of poor data quality and the necessity of domain expertise, featuring deep dives into real-world AI missteps, from a flawed enzyme classification paper to an app's mishandling of long-COVID data.

As pathology labs digitize, how do skilled practitioners learn to trust AI with patient diagnoses?

Michael Rivers discusses the critical importance of "explainable AI," keeping the human pathologist in the loop with the ability to override algorithms, and how these powerful new technologies will help address the global shortage of pathologists, ultimately bringing equitable, expert care to patients worldwide.

Join us in our first episode where we chat with Margo Georgiadis, CEO of Montai Therapeutics as we discuss what a model means in ML vs. life sciences and how she's been able to bridge the knowledge gap between these two practices for her team.

Welcome to From Model to Medicine, a podcast from KAMI Think Tank exploring how artificial intelligence is actually being applied across the life sciences industry.

Each episode, we speak with scientists, operators, and leaders working at the intersection of AI and biology, from drug discovery and clinical development to regulatory operations and beyond.

KAMI Think Tank is an educational organization that helps life science professionals learn how to use AI practically in their day-to-day work.

This podcast is about moving past the hype and understanding what it actually takes to bring AI from promising models to real impact in medicine.

We’re your hosts, Kamayani Gupta and Michelle Yi. This is From Model to Medicine.